Variation of Fe, Al, and F Substitution in Titanite (Sphene)

Titanite is an important mineral in petrochronology studies. Understanding chemical signatures of titanite from different environments can provide significant data in unraveling the complex histories recorded in their textures and compositions. Using a database of over 8500 titanite analyses from both the literature (3829) and our own data (4900), we found that the ratio of Fe/Al is useful for separating igneous titanite (Fe/Al is typically close to 1:1 and almost always > 1:2) from metamorphic titanite (Fe/Al ratio is < 1:2) with few exceptions. Volcanic titanite grains can also be separated from plutonic titanite grains due to their shorter crystallization histories with compositions clustered more tightly in terms of Fe, Al, and F. Compositions of titanite from plutonic rocks often have later metamorphic or hydrothermal overgrowths that are not found on volcanic titanite. Fe/Al ratios in titanite from silica-undersaturated volcanic and plutonic rocks are typically > 1:2 and include titanite with the highest Fe/Al ratios. Although they overlap the field for normal igneous titanite, other elements (particularly high levels of Nb and low levels of Y) allow them to be separated. In most metamorphic rocks, the Fe/Al ratio is < 1:2 except for a few metamorphic titanite grains that formed in mafic rocks. Titanite from ultrahigh pressure metamorphic rocks (eclogite facies) tend to have the lowest Fe/Al ratios, typically < 1:8. Titanite from hydrothermal and pegmatitic environments scatter widely in terms of Fe/Al even within single grains due to crystallization from fluids with highly variable compositions. Charge balancing in metamorphic, hydrothermal, and pegmatitic titanite due to Fe+3 and Al+3 substitution into the Ti+4 site is largely accomplished by the coupled substitution of F− for O−2. However, in volcanic and plutonic titanite, the charge imbalance due to Fe+3 and Al+3 substitution appears to be mainly coupled with REE+3 or Y+3 substitution into the Ca+2 site with a lesser contribution from F−.

The name titanite was settled upon as the official name of the mineral [20] and will be used in this paper to refer to all the different varieties.
Fe, Al, and F are typically the most abundant elements that substitute into the titanite structure and are reported in most published analyses. In this paper, we reported on compositional trends in titanite, with emphasis on Fe, Al, and F. The entire titanite database used in this study may be accessed as a Supplemental File associated with this publication.

Methods
Data (Supplementary Tables S1 and S2) were collected from the literature sources (3829 analyses) and supplemented with our own analyses (4900 analyses). Where particular data sets are highlighted in figures in the text, we have indicated the sources of these analyses in the figure caption or on the figure. No source is listed when we are using the entire database or when the data come from our own analyses. Analyses compiled from the literature do not, as might be expected, always include the same elements. Ca, Ti, Si, Al, Fe, and F were typically analyzed, but beyond those, the variability in elements analyzed is wide. Some caution should, therefore, be used when comparing total REEs between samples, for example. Plots of the data for individual elements have not included samples where no value for an element was reported or where the value was reported as below detection limits.
Our new analyses of titanite were made on a Cameca SX-50 electron microprobe at either the University of Utah or Brigham Young University. Data were collected on both mineral separates and thin sections at 15 kV and 20-30 nA using a spot size of 5-10 µm. Counting times were typically 10-20 s for major elements and 20-40 s for REE and minor elements. The data were reduced using a modified ZAF technique. Elements analyzed on each crystal were TAP-Si, Al, Y, Mg, Zr; LiF-Fe, La, Ce, Nd, Sm, Ta; PCO-Na, F; and PET-Ca, Ti, Mn, Nb. The following standards were used: LaPO 4 , CePO 4 , NdPO 4 , SmPO 4 , (REE standards from Smithsonian [44]), Woll (Ca, Si), Mag (Fe), Anorth (Al), MgO (Mg), Jad (Na), Taylor sphene (Ti), Spess (Mn), Yag (Y), Li 2 Nb 2 O 6 , Li 2 Ta 2 O 6 (Nb, Ta), and Taylor zircon (Zr). Standardization was checked by analyzing Taylor sphene as well as an inhouse standard titanite from an Ontario, Canada alaskite (Supplementary Table S3). The calculated atoms per formula unit was based on three total cations. Back-scattered electron (BSE) images were acquired with an ESEM FEI XL30 at Brigham Young University using an accelerating voltage of 15 kV and current of 3.2-6.4 nA with an image size of approximately 3000 × 2000 pixels.

Substitutions into Ti-Site of Titanite
Ti occurs in titanite in octahedral coordination and is commonly substituted for by Fe, Al, Nb, Ta, Mg, Zr, Sn, and to a lesser extent by V, Cr, U, and Th [1,45,46]. In some titanite, the percentage of this site being occupied by elements other than Ti can be quite high, up to 30-65% (Figure 1; see the following reports for examples [31,). The most common substitution is that of trivalent Fe and Al into the site dominated by quadrivalent Ti. This leaves a charge imbalance that is mostly accommodated by a coupled substitution of F − or OH − for the one oxygen (called O1) that is underbonded compared to other oxygens in the structure [2,4,39,[70][71][72][73][74][75]. Other significant substitutions are also known and play a role in the overall charge balancing [4]. Figure 2 shows the variability of Ti versus other Ti-site elements in titanite from different environments where titanite occurs. In volcanic rocks, about 5-18% of the site may be occupied by other elements. Volcanic rocks, however, show the tightest grouping of substitute elements into this site. In plutonic rocks, the typical levels of substitutions range between about 5 and 27%, while in some unusual granites and in metamorphic rocks, pegmatites, and hydrothermal deposits up to about 65% of the Ti site may be occupied by other elements. Figure 1. Ti (atoms per formula unit) in titanite plotted versus Fe + Al (red) and other elements likely to substitute into the Ti-site (blue). Fe + Al follow closely along a 1:1 substitution line, other elements that likely go into this site also trend along this line but at a shallower slope. The scatter of blue dots along the Ti axis is likely due to analytical difficulties in measuring low concentrations of these other elements. Data sources for this and other diagrams are given in Supplementary Table S1. Analyses that fall off the main substitution trend in Figure 2 are all from somewhat unusual titanite. In the plutonic rock data, a group of outliers with overfilled Ti sites are from peralkaline A-type granites from China and have very high Al values (up to 0.35 apfu) [69]. A second group is from a porphyry-Cu deposit in China [43] and Precambrian granitoids in South Africa [76]. In these titanite grains, the Si site is underfilled with only enough Si to account for 85-90% of the atoms needed for that site. This suggests that some element that we have assigned to the Ti site has instead substituted into the Si site-perhaps Al.
Titanite from nepheline syenites associated with lujavrites from China [77] has elevated Ti-2-4% more TiO 2 than any other titanite in the silica undersaturated category ( Figure 2). Several groups of hydrothermal titanite also have overfilled Ti sites. First, titanite from veins in the Sullivan Pb-Zn-Ag deposit in British Columbia, Canada [78] has high Fe + Al (0.25 to 0.54 apfu) and is overfilled in the Ti site similar to high Fe + Al titanite from other environments ( Figure 2). Second, titanite from veins in nepheline syenites of the Kola Peninsula [63] does not have high Fe + Al but has high Nb (0.20 to 0.36 apfu) and Zr (0.05 to 0.15 apfu). Third, other analyses with overfilled Ti sites come from the A-type Skye granite [55] that has titanite (possibly secondary) with extreme compositions (Fe up to 0.5 atoms per formula unit (apfu) as well as over 0.7 apfu of REEs) and from skarns in the Czech Republic [51,61] where titanite again has high Nb (up to 0.24 apfu) like the titanite from the Kola peninsula.
In all these cases across all environments, titanite with an overfilled Ti site appears to have high values of some element assigned normally to the site. This suggests that in these extreme conditions some element is instead substituting into the Si site, which is usually underfilled. This hypothesis warrants further study to determine if this is indeed the case.

Trends with Iron, Aluminum, and Fluorine
Nakada [39], using a more limited data set than the one presented here, observed that volcanic rocks typically have Fe/Al ratios of approximately 1:1, plutonic rocks a ratio of 1:2-1:1, metamorphic rocks less than 1:2, and that Fe typically increases as Al increases. These, as well as other trends, were examined using our expanded data set. We separated the titanite analyses into the following categories ( Figure 3): (1) Volcanic: including titanite from all types of volcanic rocks except those from silicaundersaturated rocks. (2) Plutonic: including titanite from all types of plutonic rocks except pegmatites and silica-undersaturated rocks. (3) Silica-undersaturated igneous titanite: including titanite from both volcanic and plutonic rock types. (4) Metamorphic: including titanite from all types of metamorphic rocks except those from skarns and other metasomatic and hydrothermal environments. (5) Hydrothermal: including titanite formed in skarns and other water-rich environments (like many ore deposits), and sometimes occurring as overgrowths on plutonic or metamorphic titanite. (6) Pegmatitic.
Our compilation of data broadly confirmed the observations of Nakada [39,79] for metamorphic versus igneous titanite with some exceptions, but we did not see a significant difference in the Fe/Al ratio between volcanic and plutonic titanite ( Figure 3).
Volcanic titanite: Titanite has been described as uncommon or less common in volcanic than in plutonic rocks [1,2,39,80,81]. However, titanite has been found to be quite common in some volcanic suites [82][83][84][85]. Previous studies on volcanic titanite focused on rare earth elements (REE), which partition strongly into accessory minerals such as titanite [80,82,[86][87][88][89][90]. These studies did not examine Fe and Al trends in any detail. We found that in titanite from volcanic rocks the Fe/Al ratio is > 1:2 and the vast majority of samples plot near the 1:1 line. A few data points plot at values > 2:1. These high-Fe analyses are from titanite from three samples: a Cenozoic andesite lava flow from the Basin and Range (new BYU data) and two Late Jurassic tuffs from the Morrison Formation of the Colorado Plateau [83]. We found no volcanic titanite with Fe/Al ratios < 1:2 ( Figure 3). Although volcanic titanite clusters in a limited range on the Fe versus Al plot when all analyses from the database are combined (Figure 3), analyses from individual volcanic units form even tighter clusters ( Figure 4) and typically show a positive correlation between Fe and Al. The clustering of data may be useful as an additional tool in helping to "fingerprint" and correlate volcanic ash beds and tuffs. For example, in titanite from two samples of silicic fallout tuff in the Jurassic Morrison Formation, one from northeastern Utah (DQW-21) and one from central Utah (LCM-12), Fe/Al ratios were unusually high (>2:1) ( Figure 4). Based on this and other mineralogical similarities, as well as stratigraphic position, Christiansen et al. [83] proposed that the two are likely correlative and from the same ash bed. Variations in Fe and Al within grains tend to be small in most volcanic titanite. Figure 5 shows data for three grains from the Cenozoic Lund Tuff of western Utah. There was a positive correlation between Fe and Al within each grain's group of analyses. Because of this correlation, the ratio of Fe to Al remains almost constant across a grain. Fluorine and REE, however, are quite variable. Dark zones (poor in REE) in the backscattered electron (BSE) image have high F contents, while light zones (enriched in REE) have low F concentrations ( Figure 6). Apparently in this case, F − for O −2 exchange was driven by charge balancing REE +3 substituting for Ca +2 .
Titanite grains from pyroclastic rocks are often broken with some preserved crystal edges (Figures 6 and 7). Internally, they can have quite complex zoning ( Figure 7A) or simple zoning ( Figure 7B).   Plutonic titanite: Titanite from plutonic rocks plots in the same general area as volcanic titanite with most of the data having Fe/Al ratios > 1:2 ( Figure 3). Among all the plutonic samples we examined, only a few suites of titanite analyses plotted consistently at Fe/Al ratios > 2:1. One group is from the Middle Jurassic Juniper Flats syenite (Figure 3), an A-type alkaline intrusion in the San Bernardino Mountains of southern California [92]. The low Al is the most distinctive feature. It is likely that the low Al content reflects a low aAl 2 O 3 in this nearly peralkaline sample (it has 850 ppm Zr, as well as low CaO and MgO coupled with high alkalies). Titanite from other A-type peralkaline intrusive and volcanic rocks has similar high Fe/Al ratios [69,77,[93][94][95][96][97][98][99][100][101] (Figure 8). Two titanite data sets not shown on Figure 8, which also have Fe/Al ratios > 2:1, are from a kersantite lamprophyre in the Frankenwald, Germany [102], which has about 0.28 Fe apfu compared to typical plutonic titanite which usually has > 0.1 apfu, and data from the Skye Granite, Scotland [55], which is likely secondary titanite, with Fe content of about 0.45 apfu, the highest iron content of any titanite in our database. These two data sets (seven analyses total) are not included on Figures 3 and 8 because they would extend the Fe axis by another 150% and compress the relationships in the rest of the diagram. . Fe versus Al for titanite from peralkaline igneous rocks compared with fields (90% of data) for titanite from other volcanic (red shading) and plutonic rocks (green shading). Titanite from peralkaline rocks typically has Fe/Al > 2:1. It is possible that the clump of volcanic data (red dots) that plot with Fe/Al > 2:1 is also peralkaline, but these data points were collected from altered volcanic tuffs (bentonites) from the Jurassic Morrison Formation where the original rock compositions are difficult to determine.
When compared to volcanic titanite, there is more scatter in the plutonic data set, particularly to lower Fe/Al values. This scatter to lower ratios may be due in part to overgrowths on or recrystallization of plutonic titanite by metamorphic or hydrothermal processes. There is also the possibility that some titanite of hydrothermal or metamorphic origin has been misclassified as plutonic. We identified plutonic titanite in our own data and in some published reports where it was clear that later metamorphic or hydrothermal titanite had grown around earlier formed magmatic grains. This appears to be particularly common in skarns and ore deposits with their associated plutons [31,35,57,[103][104][105], and is likely also quite common in other plutons not directly associated with ore deposits. A few examples are discussed in more detail here.
Plutonic Example 1: Titanite grains from a small pluton associated with copper skarn deposits in the Rocky Range west of Milford, Utah [106,107], appear to have plutonic cores that grew around partially resorbed Fe-Ti oxides (Figures 9 and 10). These plutonic cores have in turn been partially resorbed and then overgrown by hydrothermal titanite. This late-stage hydrothermal titanite has grown out into and between surrounding grains, particularly chloritized biotite ( Figure 10). Tulloch [108] showed that titanite is a common product of the "low-grade breakdown of biotite, being the chief depository for biotite Ti." The hydrothermal overgrowths have low Fe/Al of approximately 1:3.5, while the cores are more typical of plutonic titanite with Fe/Al of around 1:1 ( Figure 9).
The association of Fe-Ti oxides overgrown by, or associated closely with, titanite was previously observed in both plutonic and metamorphic titanite [105,[110][111][112][113][114]. The authors in most of these papers found zoned grains similar to those shown in Figure 11. Angiboust and Harlov [21] studied the replacement of ilmenite by rutile and titanite in metamorphic rocks. Their experiments showed that ilmenite, "in the investigated P-T range [0.7-1.6 GPa, 450-650 • C] . . . reacts during metamorphism above 1.2-1.4 GPa to form rutile. Below this pressure, titanite is the dominant Ti-bearing species for most granitoid compositions" [21] (p. 1696). They also noted that the position of the reaction curve is strongly influenced by whole-rock Ca concentrations. Even though this relationship with Fe-Ti oxides is found in both igneous and metamorphic titanite, in most metamorphic rocks, the titanite (see, for example, [115]), is high in Al with Fe/Al around 1:2, while in plutonic rocks the Fe/Al is around 1:1. Harlov et al. [115] related the growth of titanite around ilmenite in amphibolite-facies metapelites and amphibolites from India and Italy to hydration reactions in the presence of high fH 2 O and perhaps high fO 2 . These are likely also factors in this same relationship found in plutonic rocks.   In addition to the published examples listed above, we also observed this association of titanite that formed around Fe-Ti oxides in the Little Cottonwood stock and the Notch Peak pluton, both in Utah ( Figure 11).
Plutonic Example 2: We analyzed titanite from several Jurassic intrusions in the San Bernardino Mountains of southern California, which have been studied by Barth et al. [92], including the Rattlesnake Mountain Granite, John Bull Flat Granite, Crystal Creek Syenite, and Juniper Flats Syenite. Although the titanite compositions from each pluton cluster fairly well, there is considerably more scatter in these samples than in the volcanic rocks examined previously. Volcanic titanite typically varies less than 0.03 apfu in Fe and 0.02 apfu in Al in any given sample (Figure 4), while these plutonic titanite grains vary more than 0.03 apfu in Fe, with titanite in one sample varying by as much as 0.07 apfu (Rattlesnake Mountain) and by more than 0.03 apfu in Al up to over 0.08 apfu ( Figure 12). These larger variations in Fe and Al in plutonic titanite are probably due to a longer and more complex growth history. Figure 13 plots Ce + Nd versus Y for titanite from these California granites and compares their growth paths with titanite from several Cenozoic tuffs from the Great Basin and Colorado. The titanite in granite typically has large differences from core to mantle to rim when compared to the titanite in volcanic rocks. According to Kowallis et al. [36], this may be a general difference between volcanic titanite and plutonic titanite. Volcanic titanite is typically not as strongly zoned as plutonic titanite and lacks the irregular rims observed in many plutonic titanite grains that have low REE and Y and high F and Al concentrations.  In Figure 12, the low ratios for Fe/Al (<1:2) are due to overgrowth or replacement by hydrothermal titanite. Figure 14 shows a BSE image along with electron microprobe transects across two grains from the Rattlesnake Mountain pluton. These grains are cut by cracks along which the original titanite has either been altered or replaced. Grain B also has an area of recrystallized or overgrown titanite crossed by the transect along the bottom margin. At each crack, the titanite grains are darker and have reduced Fe/Al and Fe/F ratios. Notice that in grain A only one crack is traversed by the microprobe line. In this grain, the Fe/Al ratio increases slightly from the euhedral rim of the grain at analysis point 1 into the brighter core but does not vary much from 1:1 except around the crack where the ratio drops below 0.5. This grain also shows a typical pattern of enriched F at the rim as shown by decreasing Fe/F and Al/F ratios toward the rim of the grain, except where the crack is encountered and a sharp drop in these ratios occurs. The increase in F and drop in Fe at the crack seems most likely to have been produced by hydrothermal alteration of the titanite along the crack. This same pattern is repeated several times in grain B where cracks are more common.  In other published data sets, we found compositions that appeared more like metamorphic or hydrothermal titanite but where the authors had not specifically identified them as such.
Here are examples of some of these.
Plutonic Example 4: Titanite from plutons associated with porphyry copper deposits in the Cenozoic Red River alkaline igneous belt in southwest China [43] fall into two distinct groups based upon whether they come from mineralized or barren intrusions ( Figure 14). Titanite from mineralized intrusions in this study fall along the 1:1 line typical for magmatic titanite, while titanite from barren intrusions plots between the 1:2 and 1:4 ratio lines. We had no reason to doubt the classification as plutonic by the authors of this study. All the titanite shown in images in their paper appeared to be euhedral and have typical plutonic crystal shapes; however, the compositions of titanite from the barren intrusions are certainly unusual for plutonic titanite.
Plutonic Example 5: Another interesting occurrence comes from a study by Che et al. [31] of skarns and associated rocks, including plutonic rocks, in Canada. Plutonic (magmatic) titanite grains from the Jurassic Takomkane Batholith associated with the Boss Mountain molybdenum deposit are euhedral and plot along the 1:1 Fe/Al ratio line typical for plutonic titanite, but titanite from the Cretaceous Dublin Gulch stock associated with the Ray Gulch tungsten deposit is described as subhedral and plots between Fe/Al of 1:4 and 1:8, strongly aluminum-rich and more typical of metamorphic or hydrothermal titanite ( Figure 16). It may be that these subhedral grains in the Ray Gulch deposit are not primary magmatic titanite.
Plutonic Example 6: Gogoi et al. [116], in a study of titanite ocelli from the mafic-felsic interaction zone of the Nimchak Granite Pluton of Chotanagpur Granite Gneiss Complex of Eastern India, obtained Fe/Al ratios between 1:2 and 1:6 ( Figure 16). The Chotanagpur Granite Gneiss Complex is a Proterozoic high-grade metamorphic terrain. Even though these titanite grains come from a rock classified as plutonic, it is likely that they have a strong metamorphic overprint and should perhaps be classified as metamorphic rather than plutonic. Plutonic Example 7: Rossetti et al. [117] studied titanite from the Menderes Massif, western Turkey, closely associated with the Gediz detachment zone. They identified two groups of titanite, one of which they classified as magmatic, and a second group classified as metamorphic overgrowths ( Figure 16). Both groups of titanite have Fe/Al < 1:4 and would appear to be metamorphic. However, Rossetti et al. [117] pointed out that the cores, which they classified as magmatic, have significant differences in other elements such as the REE. This difference can be seen in a plot of REE vs. Fe (normalized to Fe/[Fe + Al]; Figure 17). The magmatic cores from Rossetti et al. [117] do plot with REE values more typical of magmatic titanite, but their Fe/Al values are not typical of most other plutonic and volcanic titanite. This is an interesting occurrence that perhaps warrants further investigation to determine why these magmatic titanite grains are so different.
Titanite grains in plutonic rocks are typically euhedral with sharp crystal faces (Figures 11, 14 and 18), although they may have late, irregular overgrowths such as those seen in Figures 14 and 18. However, titanite grains in some plutonic rocks have quite irregular shapes, suggesting resorption after initial growth ( Figure 10). Another important process in altering the chemistry of titanite is recrystallization. Garber et al. [109] documented this in some plutonic titanite grains. Dissolution-reprecipitation may have occurred along the cracks and margins of the grains shown in Figure 14 and was involved in the grains from the Notch Peak pluton, Utah, shown in Figure 19. The Notch Peak pluton grains have maintained their original sharp euhedral shape while recrystallization has created embayments and replaced large sections of the original grains with a titanite that is lower in REE and higher in Al and F (dark in the photos). These recrystallized zones cut across the earlier formed oscillatory zoning but do not appear to significantly modify the euhedral crystal shape.    The replaced portions of the crystals are darker (lower in REE) and cut across the earlier formed oscillatory zones, but the exterior shapes were nearly preserved. In (A), the large grain has a prominent dark embayment of recrystallized titanite along the lower central margin. Below this embayment are several smaller titanite crystals that have all been partially replaced. (B) A titanite grain where almost half of the original has been replaced with a distinctive low REE and high Al composition. In all the grains, the original euhedral shape has been preserved during the recrystallization process.
Titanite from silica undersaturated igneous rocks: Silica undersaturated volcanic and plutonic rocks (phonolites, foidites, nephelinites, jacapurangites, ijolites, etc.) typically have titanite with Fe/Al ratios > 1:2 ( Figure 3). It is in this group of titanite where we found the highest Fe/Al values. Wu et al. [77], in a study of sodic lujavrites from the Saima alkaline complex, northeastern China, measured Fe/Al values > 10:1 in several titanite grains. The data shown in Figure 3 [77], and for Magnet Cove, Granite Mountain, and Kaiserstuhl (this paper). Shaded area is for 90% of plutonic titanite for comparison.
Although titanite compositions from these undersaturated igneous rocks overlap with the data fields for normal volcanic and plutonic rocks when plotted as Fe versus Al (Figure 3), they are chemically distinctive in other ways. For example, they have very low concentrations of Y and high concentrations of Nb when compared to titanite from other environments. In Figure 21, we plotted Fe (normalized to Fe + Al) and Nb (normalized to Nb + Y). Most of the different genetic titanite types (volcanic, plutonic, metamorphic, and eclogitic) separate on this plot due to their characteristic Fe and Al concentrations, but they are not particularly distinctive in Nb and Y. However, titanite analyses from silica-undersaturated rocks clump at the high end of the Nb/(Nb + Y) axis, quite different from the other types. Among the titanite analyses from silica-undersaturated rocks on Figure 21, the four main data sets for this genetic type can still be seen (as they were on Figure 20 [118][119][120]) are peralkaline and their titanite has Fe/Al > 2:1. Sphene nepheline syenite and jacupirangite from Magnet Cove and the rocks of Granite Mountain are not peralkaline [77,[121][122][123] and plot at Fe/Al < 2:1. The Limberg t3 Tuff has titanite that plots at Fe/Al ≈ 2:1. The low Al activity in the peralkaline rocks may explain the low concentrations of Al in their titanite grains.
Magmatic titanite grains from both volcanic and plutonic types of silica undersaturated rocks tend to be euhedral and may display several different types of zoning, similar to other volcanic and plutonic titanite ( Figure 22). Interstitial, anhedral titanite has also been observed in plutonic rocks of this type [77], like what is observed in normal plutonic rocks.
Titanite from metamorphic rocks typically has Fe/Al < 1:2, with some data points scattering up to higher values (Figure 3). The high Al content of typical metamorphic rocks is probably controlled by the protolith. The metamorphic rocks in our compilation are overwhelming peraluminous metapelitic rocks with relatively high Al2O3 and low alkali contents (e.g., metagraywacke, metaargillite, schist, gneiss, etc.). Thus, when such aluminous rocks are metamorphosed, the titanite is also Al rich.
Titanite from high-pressure metamorphic rocks (mostly eclogites) is distinct from most other metamorphic titanite and typically has very low Fe/Al < 1:8, except in titanite with low Al (<0.15 apfu) where a few data points scatter up into higher Fe/Al (Figure 23). The abundance of Al in these high and ultrahigh pressure (UHP) rocks has been proposed as a solution to the stability of titanite where it would otherwise be unstable [2] and may be the reason for their high Al content. However, Ye et al. [131] found UHP rocks with titanite that had low Al (average~0.07 apfu). However, even in these lower Al titanite grains, the average Fe:Al ratio is still~1:8.  Another distinctive group among the metamorphic titanite analyses are those with mafic protoliths such as metabasalt, metabasanite, metapyroxenite, and amphibolite ( Figure 3). Titanite from these rocks typically has Fe/Al > 1:2, more akin to plutonic and volcanic titanite than the rest of the metamorphic titanite. This is likely due to the Fe-rich nature of the mafic protoliths.
Titanite grains in metamorphic rocks may be idoblastic to xenoblastic (euhedral to anhedral) (Figure 24). Unlike plutonic titanite grains, which typically have large (111) crystal faces giving them a characteristic diamond shape (see Figure 19), metamorphic titanite crystals are usually less angular in appearance. Hydrothermal and pegmatitic titanite: Mazdab et al. [38] noted that titanite from hydrothermal environments showed a wide compositional diversity. Our compilation confirmed this observation. Analyses from grains that grew in both hydrothermal and pegmatite environments scatter widely in terms of Fe/Al with no obvious trends in the overall data set (Figure 3). The scatter of compositions among hydrothermal titanite is probably due to crystallization from fluids with variable temperatures and compositions in a wide variety of rocks. The oxidation state also plays an important role. This can be seen clearly in Figure 25 where titanite analyses from typically oxidized Fe, Cu, and Au skarns and mineral deposits plot with Fe/Al > 1:3, while titanite from generally reduced W, Sn, and Mo skarns typically plots at Fe/Al < 1:4. Apparently, Fe +3 substitutes more readily for Ti +4 than Fe +2 . Figure 25. Fe versus Al in titanite from skarns and related ore deposits. Titanite from deposits rich in Fe, Cu, and Au (squares and red shades) plot at higher Fe/Al values (>1:3) than does titanite from deposits rich in W and Sn (circles and green shades, which have Fe/Al values (<1:4), except for a few data point exceptions. The data come from the following sources: [16,31,35,51,57,61,105,[132][133][134].
We briefly discussed in the earlier section on plutonic titanite the occurrence of secondary, hydrothermal titanite and will expand on it here. Hydrothermal titanite in plutonic rocks has been documented in several studies [42,50,[135][136][137][138]. Morad et al. [137] observed tiny lenticular titanite grains that grew along the cleavage planes in chloritized biotite. Xie et al. [42] found a similar association of secondary titanite and chlorite. Janeczek [136] showed that titanite is a common byproduct of the chloritization of Ti-bearing biotite and often produces high-Al titanite grains. We observed the same type of secondary titanite growing along chloritized biotite cleavage planes in samples from the Rocky Range pluton, Little Cottonwood stock, and Notch Peak pluton, all in Utah (Figures 10 and 18). In these Utah intrusions, the secondary hydrothermal titanite that grew along biotite/chlorite cleavage planes had Fe/Al of 1:3 to 1:5, while the magmatic titanite in these plutons had Fe/Al of around 1:1 [24].
In granitic pegmatites, crystallization occurs across a range of conditions from volatilerich magmas to vapor phase to hydrothermal fluids [139][140][141][142][143]. This poses a problem in classification. Should titanite analyses from a particular pegmatite be considered plutonic? Or perhaps hydrothermal? Or something else? Without additional data to determine what conditions might have existed during titanite crystallization, we cannot place them with certainty into a plutonic or hydrothermal category. Therefore, we chose to place any titanite analyses that have been called pegmatitic into their own category.
Because the compositions of the volatile-rich magmas and fluids that form pegmatites are quite varied, so also are the titanite grains that crystallize from them. Figure 26 shows data from several pegmatites plotted as Fe versus Al. The data scatter widely from those rich in Fe (Fe/Al > 1:1) to those rich in Al and poor in Fe (Fe/Al < 1:4). One of the unique features of titanite in pegmatites is that it may be rich in elements not normally found in other types of titanite. For example,Černý et al. [52] reported that titanite from the Maršíkov II pegmatite in northern Moravia, Czech Republic, may contain up to 21% Ta 2 O 5 and 9.5% Nb 2 O 5 . Titanite compositions have also been reported from pegmatites in Manitoba [59,67], Poland [62], Norway [64], Slovakia [144], and British Columbia [145] with elevated levels of Nb and/or Ta. Most titanite analyses from a particular sample from a hydrothermal or pegmatite environment cluster together on an Fe versus Al plot ( Figure 26). However, some do not. Included here are a few examples of the variations that may occur in single samples and even single grains of pegmatitic or hydrothermal titanite.
Hydrothermal Example 1: Anhedral titanite grains filling cracks in tantalian rutile from the Maršíkov II pegmatite [52] plot in two clumps on Figure 26, one with low Al and one with high Al.Černý et al. [52] identified three types of titanite in this pegmatite: a Na-enriched variety with low Al, a F-enriched variety with the highest Al values, and a third variety with high Al and the highest Ta/Nb ratios. All these varieties are thought to have formed during retrograde metamorphism by fluid released during dehydration reactions [52].
Hydrothermal Example 2: Even within a single grain, variations in Fe and Al can be quite extreme. Chakhmouradian [146] documented secondary deuteric titanite with a wide variety of compositions within individual grains that crystallized in carbonatites of the Afrikanda area of the Kola Peninsula, Russia. Grains had wide ranges of Fe 2 O 3 (up to 8%), Al 2 O 3 (up to 10%), ZrO 2 (up to 15%), and Nb 2 O 5 (up to 11%).
Hydrothermal Example 3: An additional example of the variations that may occur within a single hydrothermal titanite crystal can be seen in Figure 27. This crystal grew into a miarolitic cavity in an aplite located west of Milford, Utah. The cavities in this aplite have euhedral crystals of smoky quartz, fluorite, titanite, adularia, albite, and epidote [106]. The cores of titanite crystals from these miarolitic cavities have Fe/Al of 1:1, as well as LREE concentrations typical of magmatic titanite (see Figures 3 and 17 for a comparison with plutonic and volcanic titanite) even though they grew into an open cavity. First, was the growth of the core of the grain, which appears to have been euhedral with sharp crystal faces (note the sharp, straight oscillations near the margin of the interior zone on Figure 28). Subsequently, a cracking event occurred. The cracks were then annealed leaving trails of fluid inclusions and, in at least one case, a filling of slightly different titanite with a composition similar to material in the next zone that formed around the core ( Figure 28). These annealed cracks do not extend out into the outer zones of the crystal but are confined to the interior core zone. These miarolitic cavities also contain K-feldspar, albite, epidote, smoky quartz, and fluorite crystals. This titanite grain was broken out of the cavity but preserves (along top of photo) the euhedral edge that was growing into the cavity. The graphs at the right show that the dark gray rim has high Al and F as well as low Fe and LREE (La, Ce, Nd, Sm) concentrations (b). Mantling the core of the grain is a zone with wavy, irregular light and dark bands and patches. Fluid inclusions are abundant along some of the boundaries between bands in this zone. The wavy character along with the fluid inclusions suggest oscillating periods of dissolution and regrowth ( Figure 28). The analyses from this wavy zone still have Fe/Al close to 1:1 (Figure 27, labeled as 'intermediate' analyses), and variable bright and dark areas corresponding to higher and lower LREE, but with a trend to somewhat lower values toward the rim.
Finally, a dark gray outer rim overgrew these interior zones. This rim zone has a spongy appearance in the BSE image, perhaps due to fluid-or vapor-filled cavities ( Figure 28). This rim zone is characterized by high Al and F, and low Fe and LREE ( Figure 27). The Fe/Al in this rim zone is < 1:15 and ranges as low as 1:35. This relative enrichment in Al is greater than any values seen in the examples given earlier for overgrowths on plutonic titanite.

Charge Balancing
The substitution of Fe and Al into the Ti site leaves a charge imbalance that is thought to be mostly compensated for by F − or OH − substitution for the one O atom that is not bound in SiO 4 −2 groups [1,4,18,45]. Plots of Fe + Al versus F in titanite are shown in Figure 29. In all the genetic types, there is a positive correlation between Fe + Al and F. However, the correlation is much stronger in metamorphic, pegmatitic, and hydrothermal titanite than in volcanic and plutonic titanite (including silica-undersaturated types). When Fe and Al are broken out and plotted separately versus F, Al has a strong correlation with F while Fe does not (Figures 30 and 31), an observation that has been reported previously in a study of metabasites from the Canadian Shield [127]. From our data set, we can see that Al and F are strongly correlated in most titanite--except volcanic titanite ( Figure 30). The overall correlation for pegmatites is also not strong, but this is partially an artefact of the pegmatite data set, which is strongly influenced by two samples. Titanite in these two samples taken separately has fairly strong correlations between Al and F. None of the genetic types of titanite show a strong correlation between Fe and F, except the overall data for pegmatites, but again this is an artefact of the preponderance of data from two samples (Figure 31), and when they are looked at individually this correlation breaks down.  In volcanic and plutonic titanite, it appears that the charge imbalance created by the substitution of Fe is largely compensated for by +3 ions (REEs, Y) substituting into the Ca site as seen by the strong correlations between Fe and REE + Y ( Figure 32). Titanite from silica-undersaturated rocks, however, shows no correlation between Fe and either F or REE + Y (Figures 31 and 33). Titanite from these rocks, particularly those from the Limberg Tuff of the Kaiserstuhl complex, typically has high Nb (up to 0.10 atoms pfu), likely as Nb +5 substituting for Ti +4 . The excess charge on this ion would help to balance any deficiency created by Fe +3 in the Ti +4 site ( Figure 34). Metamorphic titanite, including that from eclogites, has a strong correlation between Al and F, but little to no correlation between Fe and REE + Y (Figures 30 and 33), showing that charge balancing in metamorphic titanite is largely accomplished by the coupled substitution of Al +3 and F − .   The other factors that we have not taken into consideration here, in relation to charge balancing of the Ti site, are the contributions from OH − and the oxidation state of Fe, which occurs in titanite as both Fe +3 and Fe +2 [147][148][149][150][151].

Discussion
Titanite is an important mineral in petrogenetic and petrochronologic studies where an understanding of its chemistry, crystallization history, and element diffusion rates is critical in making sense of the histories it records [2,4,38,109]. We demonstrated that Fe, Al, and F, secondary elements typically analyzed in titanite, can be used to identify different environments where titanite forms. Other elements, particularly REEs, Nb, and Y, are also helpful. From our large database of chemical analyses of titanite, we demonstrated the following: (1) In all environments, the principal substitutions in the Ti +4 site of titanite are Fe +3 and Al +3 . (2) The atomic ratio of Fe/Al in titanite from both volcanic and plutonic rocks is typically close to 1:1 and almost always > 1:2. (3) Volcanic titanite compositions typically cluster more tightly in terms of Fe, Al, and F than do titanite compositions from any other environment. Analyses of volcanic titanite from individual samples cluster even more tightly than the overall volcanic group. Given the wide range of potential compositions found in other environments, this suggests that titanite in volcanic rocks typically crystallized from magmas with limited ranges of Fe and Al. (4) Plutonic titanite typically clusters more loosely than does volcanic titanite in terms of Fe, Al, and F, suggesting a wider range of compositions for the magmas from which they crystallized. Some of the broader spread may be due to later metamorphic or hydrothermal overgrowths lumped in with the plutonic data that do not typically occur on volcanic titanite. (5) Unlike their Fe, Al, and F compositions, volcanic and plutonic titanite have the largest spread of REE and Y contents when compared to titanite from other environments. (6) Most primary volcanic and plutonic titanite grains are euhedral and have prominent (111) crystal faces, giving them their typical wedge or diamond shape while metamorphic titanite grains are typically less angular with a podlike shape. (7) Fe/Al ratios in titanite from silica-undersaturated volcanic and plutonic rocks are typically > 1:2. Although they overlap the field for normal volcanic and plutonic titanite, other elements (particularly high levels of Nb and low levels of Y) separate out titanite from these rocks. (8) Peralkaline rocks include titanite with the highest Fe/Al (typically > 2:1 and often much higher). (9) In most metamorphic rocks, the Fe/Al ratio in titanite is < 1:2 except for a few that have mafic and Fe-rich protoliths such as metabasalt or metabasite. (10) Titanite from high pressure metamorphic rocks (eclogites, blue schists, etc.) tends to have the lowest Fe/Al ratios, typically < 1:8. Apparently, Al-rich titanite is stable at higher pressures. However, some pegmatitic and hydrothermal titanite, including authigenic grains, also has Fe/Al in this low range. (11) Titanite from hydrothermal and pegmatitic environments scatters widely in terms of Fe/Al ratios even within single grains. This is likely due to crystallization from fluids with more variable compositions than is found in magmas and with a wider variety of oxidation conditions as seen particularly in the development of titanite in skarns. W-Mo skarns produce Al-rich titanite and magnetite skarns are dominated by Fe-rich titanite. Fluid composition and temperature in a hydrothermal environment may also change markedly due to a different water source flooding into the environment where the titanite crystals are forming, producing crystals with internally variable compositions. (12) Charge balance in metamorphic, hydrothermal, and pegmatitic titanite due to Fe +3 and Al +3 substitution into the Ti +4 site is largely accomplished by the coupled substitution of F − for O −2 . However, in volcanic and plutonic titanite the charge imbalance due to Fe +3 and Al +3 substitution appears to be mainly coupled with REE +3 or Y +3 substitution into the Ca +2 site with a more minor contribution from F − substitution.
(13) High quality BSE photos of titanite grains are critical in observing relationships and interpreting crystallization histories. Brightness variations correspond mainly with variations in REE concentrations, which typically anticorrelate with F and Al concentrations, particularly in igneous titanite.

Conclusions
The database of titanite analyses included with this paper provides researchers with information that will help as they use this interesting mineral to unravel complex geologic processes across a wide range of environments. The most useful elements in distinguishing titanites from different environments appear to be Fe, Al, F, Nb, and REE. BSE photos are an important tool in observing compositional variations. Used along with U/Pb ages from titanite, compositions may provide additional useful data for provenance studies of clastic sediments.